Manuscript received May 28, 2024; revised July 15, 2024; accepted August 6, 2024; published December 25, 2024.

Abstract—The detection of bleeding in the gastrointestinal tract is a critical use of Wireless Capsule Endoscopy (WCE), a valuable diagnostic tool for examining the whole gastrointestinal tract, particularly the hard to reach small intestine. However, manually analyzing a large volume of WCE images is time-consuming, burdensome, and susceptible to human errors. The purpose of this research paper is to develop and assess a Convolutional Neural Network (CNN) model that can automatically analyze and classify WCE images to detect bleeding. This model is projected on a hardware (Raspberry Pi 4 model B) platform to evaluate its classification process practically. A database consisting of 892 WCE images is employed to train and evaluate the model in terms of the metrics such as accuracy, precision, recall, and F1-Score. Two dataset cases are considered (one with 70% & 30% and the other with 80% & 20%) for training and validation, respectively. The findings exhibit promising results, with a macro-average precision, recall, and F1-Score of (0.98611, 0.98438, and 0.98502) for Case 1, (0.99474, 0.99412, and 0.99440) for Case 2, respectively. Furthermore, the suggested model obtains accuracies of 0.98507 and 0.99441 for Case 1 and Case 2 respectively. Finally, the hardware of the developed model is tested with 12 s duration video (which has 56 images) and classified all the images (11 normal 45 abnormal) within this video correctly.

Keywords—wireless capsule endoscopy, bleeding, convolutional neural network, deep learning, small intestine

Cite: Ziyad k. Farej, Amer F. Sheet, and Noora M. Sheet, "On the Classification Accuracy of Wireless Capsule Endoscopy Images for Small Intestine Bleeding Detection Using Convolutional Neural Network," Journal of Image and Graphics, Vol. 12, No. 4, pp. 450-457, 2024.

Copyright © 2024 by the authors. This is an open access article distributed under the Creative Commons Attribution License (CC BY-NC-ND 4.0), which permits use, distribution and reproduction in any medium, provided that the article is properly cited, the use is non-commercial and no modifications or adaptations are made.